DAPR

DAPR数据集是由达姆施塔特工业大学的无处不在的知识处理实验室创建，旨在评估文档感知段落检索任务。该数据集包含从多个异构领域收集的5个子数据集，包括MS MARCO, Natural Questions, MIRACL, Genomics和ConditionalQA。这些数据集用于评估和比较不同的检索系统，特别是在需要文档上下文理解的复杂查询场景中的表现。DAPR数据集的创建过程涉及对原始数据的预处理和格式统一，以确保评估的一致性和准确性。该数据集主要应用于信息检索领域，特别是针对长文档中的相关段落检索，旨在解决现有检索系统在处理长文档时面临的挑战。

The DAPR dataset was developed by the Lab for Ubiquitous Knowledge Processing (UKP Lab) at Technische Universität Darmstadt, with the primary goal of evaluating document-aware passage retrieval tasks. It consists of five sub-datasets collected from multiple heterogeneous domains, including MS MARCO, Natural Questions, MIRACL, Genomics, and ConditionalQA. These sub-datasets are used to evaluate and compare the performance of different retrieval systems, particularly in complex query scenarios that require contextual understanding of documents. The creation process of the DAPR dataset involves preprocessing the raw data and unifying their formats to ensure consistency and accuracy in evaluations. This dataset is mainly applied in the field of information retrieval, especially for relevant passage retrieval from long documents, aiming to address the challenges faced by existing retrieval systems when processing lengthy documents.